maintenance/language/generateCollationData.php

   1 <?php
   2 /**
   3  * Maintenance script to generate first letter data files for Collation.php.
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18  * http://www.gnu.org/copyleft/gpl.html
  19  *
  20  * @file
  21  * @ingroup MaintenanceLanguage
  22  */
  23
  24 require_once __DIR__ . '/../Maintenance.php';
  25
  26 /**
  27  * Generate first letter data files for Collation.php
  28  *
  29  * @ingroup MaintenanceLanguage
  30  */
  31 class GenerateCollationData extends Maintenance {
  32         /** The directory with source data files in it */
  33         public $dataDir;
  34
  35         /** The primary weights, indexed by codepoint */
  36         public $weights;
  37
  38         /**
  39          * A hashtable keyed by codepoint, where presence indicates that a character
  40          * has a decomposition mapping. This makes it non-preferred for group header
  41          * selection.
  42          */
  43         public $mappedChars;
  44
  45         public $debugOutFile;
  46
  47         /**
  48          * Important tertiary weights from UTS #10 section 7.2
  49          */
  50         const NORMAL_UPPERCASE = 0x08;
  51         const NORMAL_HIRAGANA = 0x0E;
  52
  53         public function __construct() {
  54                 parent::__construct();
  55                 $this->addOption( 'data-dir', 'A directory on the local filesystem ' .
  56                         'containing allkeys.txt and ucd.all.grouped.xml from unicode.org',
  57                         false, true );
  58                 $this->addOption( 'debug-output', 'Filename for sending debug output to',
  59                         false, true );
  60         }
  61
  62         public function execute() {
  63                 $this->dataDir = $this->getOption( 'data-dir', '.' );
  64
  65                 $allkeysPresent = file_exists( "{$this->dataDir}/allkeys.txt" );
  66                 $ucdallPresent = file_exists( "{$this->dataDir}/ucd.all.grouped.xml" );
  67
  68                 // As of January 2013, these links work for all versions of Unicode
  69                 // between 5.1 and 6.2, inclusive.
  70                 $allkeysURL = "http://www.unicode.org/Public/UCA/<Unicode version>/allkeys.txt";
  71                 $ucdallURL = "http://www.unicode.org/Public/<Unicode version>/ucdxml/ucd.all.grouped.zip";
  72
  73                 if ( !$allkeysPresent || !$ucdallPresent ) {
  74                         $icuVersion = IcuCollation::getICUVersion();
  75                         $unicodeVersion = IcuCollation::getUnicodeVersionForICU();
  76
  77                         $error = "";
  78
  79                         if ( !$allkeysPresent ) {
  80                                 $error .= "Unable to find allkeys.txt. "
  81                                         . "Download it and specify its location with --data-dir=<DIR>. "
  82                                         . "\n\n";
  83                         }
  84                         if ( !$ucdallPresent ) {
  85                                 $error .= "Unable to find ucd.all.grouped.xml. "
  86                                         . "Download it, unzip, and specify its location with --data-dir=<DIR>. "
  87                                         . "\n\n";
  88                         }
  89
  90                         $versionKnown = false;
  91                         if ( !$icuVersion ) {
  92                                 // Unknown version - either very old intl,
  93                                 // or PHP < 5.3.7 which does not expose this information
  94                                 $error .= "As MediaWiki could not determine the version of ICU library used by your PHP's "
  95                                         . "intl extension it can't suggest which file version to download. "
  96                                         . "This can be caused by running a very old version of intl or PHP < 5.3.7. "
  97                                         . "If you are sure everything is all right, find out the ICU version "
  98                                         . "by running phpinfo(), check what is the Unicode version it is using "
  99                                         . "at http://site.icu-project.org/download, then try finding appropriate data file(s) at:";
 100                         } elseif ( version_compare( $icuVersion, "4.0", "<" ) ) {
 101                                 // Extra old version
 102                                 $error .= "You are using outdated version of ICU ($icuVersion), intended for "
 103                                         . ( $unicodeVersion ? "Unicode $unicodeVersion" : "an unknown version of Unicode" )
 104                                         . "; this file might not be avalaible for it, and it's not supported by MediaWiki. "
 105                                         . " You are on your own; consider upgrading PHP's intl extension or try "
 106                                         . "one of the files available at:";
 107                         } elseif ( version_compare( $icuVersion, "51.0", ">=" ) ) {
 108                                 // Extra recent version
 109                                 $error .= "You are using ICU $icuVersion, released after this script was last updated. "
 110                                         . "Check what is the Unicode version it is using at http://site.icu-project.org/download . "
 111                                         . "It can't be guaranteed everything will work, but appropriate file(s) should "
 112                                         . "be available at:";
 113                         } else {
 114                                 // ICU 4.0 to 50.x
 115                                 $versionKnown = true;
 116                                 $error .= "You are using ICU $icuVersion, intended for "
 117                                         . ( $unicodeVersion ? "Unicode $unicodeVersion" : "an unknown version of Unicode" )
 118                                         . ". Appropriate file(s) should be available at:";
 119                         }
 120                         $error .= "\n";
 121
 122                         if ( $versionKnown && $unicodeVersion ) {
 123                                 $allkeysURL = str_replace( "<Unicode version>", "$unicodeVersion.0", $allkeysURL );
 124                                 $ucdallURL = str_replace( "<Unicode version>", "$unicodeVersion.0", $ucdallURL );
 125                         }
 126
 127                         if ( !$allkeysPresent ) {
 128                                 $error .= "* $allkeysURL\n";
 129                         }
 130                         if ( !$ucdallPresent ) {
 131                                 $error .= "* $ucdallURL\n";
 132                         }
 133
 134                         $this->fatalError( $error );
 135                 }
 136
 137                 $debugOutFileName = $this->getOption( 'debug-output' );
 138                 if ( $debugOutFileName ) {
 139                         $this->debugOutFile = fopen( $debugOutFileName, 'w' );
 140                         if ( !$this->debugOutFile ) {
 141                                 $this->fatalError( "Unable to open debug output file for writing" );
 142                         }
 143                 }
 144                 $this->loadUcd();
 145                 $this->generateFirstChars();
 146         }
 147
 148         function loadUcd() {
 149                 $uxr = new UcdXmlReader( "{$this->dataDir}/ucd.all.grouped.xml" );
 150                 $uxr->readChars( [ $this, 'charCallback' ] );
 151         }
 152
 153         function charCallback( $data ) {
 154                 // Skip non-printable characters,
 155                 // but do not skip a normal space (U+0020) since
 156                 // people like to use that as a fake no header symbol.
 157                 $category = substr( $data['gc'], 0, 1 );
 158                 if ( strpos( 'LNPS', $category ) === false
 159                         && $data['cp'] !== '0020'
 160                 ) {
 161                         return;
 162                 }
 163                 $cp = hexdec( $data['cp'] );
 164
 165                 // Skip the CJK ideograph blocks, as an optimisation measure.
 166                 // UCA doesn't sort them properly anyway, without tailoring.
 167                 if ( IcuCollation::isCjk( $cp ) ) {
 168                         return;
 169                 }
 170
 171                 // Skip the composed Hangul syllables, we will use the bare Jamo
 172                 // as first letters
 173                 if ( $data['block'] == 'Hangul Syllables' ) {
 174                         return;
 175                 }
 176
 177                 // Calculate implicit weight per UTS #10 v6.0.0, sec 7.1.3
 178                 if ( $data['UIdeo'] === 'Y' ) {
 179                         if ( $data['block'] == 'CJK Unified Ideographs'
 180                                 || $data['block'] == 'CJK Compatibility Ideographs'
 181                         ) {
 182                                 $base = 0xFB40;
 183                         } else {
 184                                 $base = 0xFB80;
 185                         }
 186                 } else {
 187                         $base = 0xFBC0;
 188                 }
 189                 $a = $base + ( $cp >> 15 );
 190                 $b = ( $cp & 0x7fff ) | 0x8000;
 191
 192                 $this->weights[$cp] = sprintf( ".%04X.%04X", $a, $b );
 193
 194                 if ( $data['dm'] !== '#' ) {
 195                         $this->mappedChars[$cp] = true;
 196                 }
 197
 198                 if ( $cp % 4096 == 0 ) {
 199                         print "{$data['cp']}\n";
 200                 }
 201         }
 202
 203         function generateFirstChars() {
 204                 $file = fopen( "{$this->dataDir}/allkeys.txt", 'r' );
 205                 if ( !$file ) {
 206                         $this->fatalError( "Unable to open allkeys.txt" );
 207                 }
 208                 global $IP;
 209                 $outFile = fopen( "$IP/serialized/first-letters-root.ser", 'w' );
 210                 if ( !$outFile ) {
 211                         $this->fatalError( "Unable to open output file first-letters-root.ser" );
 212                 }
 213
 214                 $goodTertiaryChars = [];
 215
 216                 // For each character with an entry in allkeys.txt, overwrite the implicit
 217                 // entry in $this->weights that came from the UCD.
 218                 // Also gather a list of tertiary weights, for use in selecting the group header
 219                 while ( false !== ( $line = fgets( $file ) ) ) {
 220                         // We're only interested in single-character weights, pick them out with a regex
 221                         $line = trim( $line );
 222                         if ( !preg_match( '/^([0-9A-F]+)\s*;\s*([^#]*)/', $line, $m ) ) {
 223                                 continue;
 224                         }
 225
 226                         $cp = hexdec( $m[1] );
 227                         $allWeights = trim( $m[2] );
 228                         $primary = '';
 229                         $tertiary = '';
 230
 231                         if ( !isset( $this->weights[$cp] ) ) {
 232                                 // Non-printable, ignore
 233                                 continue;
 234                         }
 235                         foreach ( StringUtils::explode( '[', $allWeights ) as $weightStr ) {
 236                                 preg_match_all( '/[*.]([0-9A-F]+)/', $weightStr, $m );
 237                                 if ( !empty( $m[1] ) ) {
 238                                         if ( $m[1][0] !== '0000' ) {
 239                                                 $primary .= '.' . $m[1][0];
 240                                         }
 241                                         if ( $m[1][2] !== '0000' ) {
 242                                                 $tertiary .= '.' . $m[1][2];
 243                                         }
 244                                 }
 245                         }
 246                         $this->weights[$cp] = $primary;
 247                         if ( $tertiary === '.0008'
 248                                 || $tertiary === '.000E'
 249                         ) {
 250                                 $goodTertiaryChars[$cp] = true;
 251                         }
 252                 }
 253                 fclose( $file );
 254
 255                 // Identify groups of characters with the same primary weight
 256                 $this->groups = [];
 257                 asort( $this->weights, SORT_STRING );
 258                 $prevWeight = reset( $this->weights );
 259                 $group = [];
 260                 foreach ( $this->weights as $cp => $weight ) {
 261                         if ( $weight !== $prevWeight ) {
 262                                 $this->groups[$prevWeight] = $group;
 263                                 $prevWeight = $weight;
 264                                 if ( isset( $this->groups[$weight] ) ) {
 265                                         $group = $this->groups[$weight];
 266                                 } else {
 267                                         $group = [];
 268                                 }
 269                         }
 270                         $group[] = $cp;
 271                 }
 272                 if ( $group ) {
 273                         $this->groups[$prevWeight] = $group;
 274                 }
 275
 276                 // If one character has a given primary weight sequence, and a second
 277                 // character has a longer primary weight sequence with an initial
 278                 // portion equal to the first character, then remove the second
 279                 // character. This avoids having characters like U+A732 (double A)
 280                 // polluting the basic latin sort area.
 281
 282                 foreach ( $this->groups as $weight => $group ) {
 283                         if ( preg_match( '/(\.[0-9A-F]*)\./', $weight, $m ) ) {
 284                                 if ( isset( $this->groups[$m[1]] ) ) {
 285                                         unset( $this->groups[$weight] );
 286                                 }
 287                         }
 288                 }
 289
 290                 ksort( $this->groups, SORT_STRING );
 291
 292                 // Identify the header character in each group
 293                 $headerChars = [];
 294                 $prevChar = "\000";
 295                 $tertiaryCollator = new Collator( 'root' );
 296                 $primaryCollator = new Collator( 'root' );
 297                 $primaryCollator->setStrength( Collator::PRIMARY );
 298                 $numOutOfOrder = 0;
 299                 foreach ( $this->groups as $weight => $group ) {
 300                         $uncomposedChars = [];
 301                         $goodChars = [];
 302                         foreach ( $group as $cp ) {
 303                                 if ( isset( $goodTertiaryChars[$cp] ) ) {
 304                                         $goodChars[] = $cp;
 305                                 }
 306                                 if ( !isset( $this->mappedChars[$cp] ) ) {
 307                                         $uncomposedChars[] = $cp;
 308                                 }
 309                         }
 310                         $x = array_intersect( $goodChars, $uncomposedChars );
 311                         if ( !$x ) {
 312                                 $x = $uncomposedChars;
 313                                 if ( !$x ) {
 314                                         $x = $group;
 315                                 }
 316                         }
 317
 318                         // Use ICU to pick the lowest sorting character in the selection
 319                         $tertiaryCollator->sort( $x );
 320                         $cp = $x[0];
 321
 322                         $char = UtfNormal\Utils::codepointToUtf8( $cp );
 323                         $headerChars[] = $char;
 324                         if ( $primaryCollator->compare( $char, $prevChar ) <= 0 ) {
 325                                 $numOutOfOrder++;
 326                         }
 327                         $prevChar = $char;
 328
 329                         if ( $this->debugOutFile ) {
 330                                 fwrite( $this->debugOutFile, sprintf( "%05X %s %s (%s)\n", $cp, $weight, $char,
 331                                         implode( ' ', array_map( 'UtfNormal\Utils::codepointToUtf8', $group ) ) ) );
 332                         }
 333                 }
 334
 335                 print "Out of order: $numOutOfOrder / " . count( $headerChars ) . "\n";
 336
 337                 fwrite( $outFile, serialize( $headerChars ) );
 338         }
 339 }
 340
 341 class UcdXmlReader {
 342         public $fileName;
 343         public $callback;
 344         public $groupAttrs;
 345         public $xml;
 346         public $blocks = [];
 347         public $currentBlock;
 348
 349         function __construct( $fileName ) {
 350                 $this->fileName = $fileName;
 351         }
 352
 353         public function readChars( $callback ) {
 354                 $this->getBlocks();
 355                 $this->currentBlock = reset( $this->blocks );
 356                 $xml = $this->open();
 357                 $this->callback = $callback;
 358
 359                 while ( $xml->name !== 'repertoire' && $xml->next() );
 360
 361                 while ( $xml->read() ) {
 362                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 363                                 if ( $xml->name === 'group' ) {
 364                                         $this->groupAttrs = $this->readAttributes();
 365                                 } elseif ( $xml->name === 'char' ) {
 366                                         $this->handleChar();
 367                                 }
 368                         } elseif ( $xml->nodeType === XMLReader::END_ELEMENT ) {
 369                                 if ( $xml->name === 'group' ) {
 370                                         $this->groupAttrs = [];
 371                                 }
 372                         }
 373                 }
 374                 $xml->close();
 375         }
 376
 377         protected function open() {
 378                 $this->xml = new XMLReader;
 379                 $this->xml->open( $this->fileName );
 380                 if ( !$this->xml ) {
 381                         throw new MWException( __METHOD__ . ": unable to open {$this->fileName}" );
 382                 }
 383                 while ( $this->xml->name !== 'ucd' && $this->xml->read() );
 384                 $this->xml->read();
 385
 386                 return $this->xml;
 387         }
 388
 389         /**
 390          * Read the attributes of the current element node and return them
 391          * as an array
 392          * @return array
 393          */
 394         protected function readAttributes() {
 395                 $attrs = [];
 396                 while ( $this->xml->moveToNextAttribute() ) {
 397                         $attrs[$this->xml->name] = $this->xml->value;
 398                 }
 399
 400                 return $attrs;
 401         }
 402
 403         protected function handleChar() {
 404                 $attrs = $this->readAttributes() + $this->groupAttrs;
 405                 if ( isset( $attrs['cp'] ) ) {
 406                         $first = $last = hexdec( $attrs['cp'] );
 407                 } else {
 408                         $first = hexdec( $attrs['first-cp'] );
 409                         $last = hexdec( $attrs['last-cp'] );
 410                         unset( $attrs['first-cp'] );
 411                         unset( $attrs['last-cp'] );
 412                 }
 413
 414                 for ( $cp = $first; $cp <= $last; $cp++ ) {
 415                         $hexCp = sprintf( "%04X", $cp );
 416                         foreach ( [ 'na', 'na1' ] as $nameProp ) {
 417                                 if ( isset( $attrs[$nameProp] ) ) {
 418                                         $attrs[$nameProp] = str_replace( '#', $hexCp, $attrs[$nameProp] );
 419                                 }
 420                         }
 421
 422                         while ( $this->currentBlock ) {
 423                                 if ( $cp < $this->currentBlock[0] ) {
 424                                         break;
 425                                 } elseif ( $cp <= $this->currentBlock[1] ) {
 426                                         $attrs['block'] = key( $this->blocks );
 427                                         break;
 428                                 } else {
 429                                         $this->currentBlock = next( $this->blocks );
 430                                 }
 431                         }
 432
 433                         $attrs['cp'] = $hexCp;
 434                         call_user_func( $this->callback, $attrs );
 435                 }
 436         }
 437
 438         public function getBlocks() {
 439                 if ( $this->blocks ) {
 440                         return $this->blocks;
 441                 }
 442
 443                 $xml = $this->open();
 444                 while ( $xml->name !== 'blocks' && $xml->read() );
 445
 446                 while ( $xml->read() ) {
 447                         if ( $xml->nodeType == XMLReader::ELEMENT ) {
 448                                 if ( $xml->name === 'block' ) {
 449                                         $attrs = $this->readAttributes();
 450                                         $first = hexdec( $attrs['first-cp'] );
 451                                         $last = hexdec( $attrs['last-cp'] );
 452                                         $this->blocks[$attrs['name']] = [ $first, $last ];
 453                                 }
 454                         }
 455                 }
 456                 $xml->close();
 457
 458                 return $this->blocks;
 459         }
 460 }
 461
 462 $maintClass = GenerateCollationData::class;
 463 require_once RUN_MAINTENANCE_IF_MAIN;